数据挖掘显示MOOCs问题重重

#研究分享#【数据挖掘显示MOOCs问题重重】本月初,宾州大学研究者报告说该校的在线课程已归于失败,仅约50%的注册者学习过课程,而最终完成率仅4%。普林斯顿大学研究者通过对课程论坛的数据挖掘发现:整个课程自始至终的参与度都在下降,50%的注册者在论坛上发帖不超过2次;论坛30%多的帖子都是与课程无关的闲谈,真正有价值的帖子被淹没,并且教师的参与讨论使课程参与度不升反降。

Data Mining Exposes Embarrassing Problems for Massive Open Online Courses

Not only does student participation decline dramatically throughout the new generation of Web-based courses, but the involvement of teachers in online discussions makes it worse.


It wasn’t so long ago that the excitement surrounding online education reached fever pitch. Various researchers offering free online versions of their university classes found they could attract vast audiences of high quality students from all over the world. The obvious next step was to offer far more of these online classes.

That started a rapid trend and various organisations sprung up to offer online versions of university-level courses that anyone with an Internet connection could sign up for. The highest profile of these are organisations such as Coursera, Udacity, and edX.

But this new golden age of education has rapidly lost its lustre. Earlier this month, researchers at the University of Pennsylvania reported that the online classes it offered had failed miserably. Only about half of the students who registered ever viewed a lecture and only 4 percent completed a course.

That’s prompted some soul-searching among those who have championed this brave new world of education. The questions that urgently need answering are: what’s gone wrong and how can it be fixed?

Today, Christopher Brinton at Princeton University and a few pals offer their view. These guys have studied the behaviour in online discussion forums of over 100,000 students taking massive open online courses (or MOOCs).

And they have depressing news. They say that participation falls precipitously and continuously throughout a course and that almost half of registered students never post more than twice to the forums. What’s more, the participation of a teacher doesn’t improve matters. Indeed, they say there is some evidence that a teacher’s participation in an online discussion actually increases the rate of decline.

Brinton and co studied the discussion threads associated with 73 courses offered by Coursera. These involved 115,000 students who wrote over 800,000 posts in 170,000 different threads. The team then plotted how the volume of discussion varied through the course and what factors correlate with this decline.

Brinton and co say they’ve found various correlations with the drop. One of these is the amount of peer-graded homework on the course, a factor which moderately increases the rate of decline. More worrying is the discovery that teacher involvement in a thread seems to accelerate the decline (although it also increases the number of posts).

Just how this can be reversed isn’t clear but one potential avenue is to improve the learning experience by making the most valuable posts in discussions more easy to find.

Brinton and co say that posts fall into three categories. The first is small talk, student introductions and the like, that are of little use in completing the course. The second is about course logistics such as when to file homework. And the final category is course-specific questions which are the most useful for students.

The problem is that the useful posts are drowned out by the others, particularly the small talk. “For humanities and social sciences courses, on average more than 30% of the threads are classified as small-talk even long after a course is launched,” say Brinton and co. “Small-talk is a major source of information overload in the forums.”

To help combat this problem, Brinton and co have developed an automated system that spots small talk and filters it out of the firehose. That should help students focus on the useful posts and enhance the learning experience.

Whether that will improve the failing metrics for massive open online courses isn’t clear. But there’s clearly scope for more work to identify the reasons why the courses fail to work for so many students. And if free online education is to get some of lustre back, this work will have to be done quickly.

Ref: arxiv.org/abs/1312.2159: Learning about social learning in MOOCs: From statistical analysis to generative model

文章来源:MIT TechReview

文章链接:http://www.technologyreview.com/view/522816/data-mining-exposes-embarrassing-problems-for-massive-open-online-courses/


1 条评论

  1. zhangye说道:

    感谢分享!http://weibo.com/1711479641/Ape9Fq2Rq?mod=weibotime



无觅相关文章插件